Mental Framing Effects in Dynamic Portfolio Choice

Abstract

Experimental evidence is presented for systematic decision errors in dynamic portfolio choice tasks within a large-scale choice experiment. Participants were asked to create a contingency plan for all possible scenarios within a binomial tree model. The quality of the chosen plans critically depends on the dynamics of the probability distribution of investment returns. When returns are independent and identically distributed, most plans are close to optimal for plausible risk preferences. In contrast, when the probability distribution is dynamic, a very large majority of the plans are inefficient, even under First-degree Stochastic Dominance. The chosen allocations appear insensitive to temporal variation in probabilities, consistent with myopic loss aversion relative to a sticky stochastic reference point. Decision quality significantly improves when the task is reduced to a discrete choice from a small number of static probability distributions, which represent the final payoffs of optimal plans for a variety of standard utility functions, in addition to the participant's original plan. The results support the importance of problem framing for dynamic choice tasks and the potential effectiveness of a libertarian paternalistic approach to framing.

Keywords: choice experiment; risky choice; dynamic problem; framing effects; behavioral finance

JEL Classification: D03, D81, D92, G02, G11